A total of 1,030 females affected by idiopathic POI were recruited at the Center for Reproductive Medicine, Shandong University. The idiopathic POI was diagnosed by 1) primary or secondary amenorrhea or oligomenorrhea before 40 years of age; and 2) at least twice serum FSH >25 IU/L at an internal of 4–6 weeks. Patients with chromosomal abnormalities, ovarian surgery, chemo/radiotherapy, or autoimmune disease that might cause POI (such as systemic lupus erythematosus, Sjogren’s syndrome, rheumatoid arthritis, and autoimmune thyroiditis) were excluded.

Genomic DNA of the 1,030 patients was extracted from peripheral blood leukocytes using DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany). Exome sequences were captured with SureSelect Target Enrichment system for Illumina Paired-End Sequencing Library (Agilent Technologies, Santa Clara, CA, USA), and DNA sequencing was performed on Illumina HiSeq platform (Illumina HiSeq, San Diego, CA, USA). Reads were mapped to the hg19 reference genome, and variants were called and annotated using Genome Analysis Toolkit (GATK), ANNOVAR, and custom pipelines. The variants of HSF2BP were screened using the following strategies: 1) alter protein sequence (nonsense, missense, splicing-site variant, and coding indels); 2) allele frequencies <0.001 in the public databases 1000 Genomes, ESP6500, ExAC, and gnomAD; 3) variants predicted as deleterious variations by more than half of the software (SIFT, PolyPhen-2, MutationTaster, MutationAssessor, PROVEAN, and M-CAP); and 4) follow recessive inheritance pattern (homozygous or compound heterozygous). Then the variants were classified as pathogenic (P), likely pathogenic (LP), or variations of uncertain significance (VUS) according to the guidelines proposed by the American College of Medical Genetics and Genomics (ACMG) (Richards et al., 2015).

The wild-type HSF2BP-FLAG plasmids were generated by inserting human HSF2BP cDNA tagged with FLAG into pcDNA3.1 vector. One non-pathogenic single-nucleotide polymorphism (SNP) with relatively high allele frequency in gnomAD was chosen as a negative control. The mutant and polymorphic HSF2BP-FLAG plasmids were generated using QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies) according to manufacturer’s instructions, with the primers listed in Supplementary Table S1 and confirmed by Sanger sequencing.

HeLa cells (human cervix carcinoma cell line) were cultured in Dulbecco’s modified Eagle’s medium/high glucose (Gibco, Grand Island, NY, USA) medium supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin–streptomycin (Gibco). KGN cells (human granulosa cell line) were cultured in Dulbecco’s modified Eagle’s medium/F-12 (Gibco) medium supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin. The cells were cultured at 37°C with 5% CO₂.

HeLa cells and KGN cells were cultured in 24-well plates and transiently transfected with wild-type, mutant, and polymorphic HSF2BP-FLAG plasmids with the use of Lipofectamine 3000 (Invitrogen, Carlsbad, CA, United States). After being cultured for 48 h, the cells were divided into three groups: without treatment, treated with etoposide (ETO; 5 μg/ml, Solarbio, Beijing, China) for 1 h, and treated with mitomycin (MMC; 500 ng/ml, Sigma-Aldrich, St. Louis, MO, United States) for 8 h. Then, the treated cells were cultured with a normal medium and recovered for a specific time at 37°C. Then, the cells were fixed with 4% paraformaldehyde (Solarbio) for 20 min at room temperature. Then the slides were permeabilized with 0.3% Triton X-100, blocked with 10% bovine albumin for 1 h, incubated with anti-FLAG antibody (1:500 dilution, CST14793, Cell Signaling Technology, Danvers, MA, United States) at 4°C overnight, followed with incubation with fluorescent secondary antibody for 1 h, and then treated with antifade mounting medium with DAPI (Beyotime, Shanghai, China). Intervening phosphate-buffered saline (PBS) washes were performed after incubation when necessary. Images were captured with an Olympus (Tokyo, Japan) BX61 microscope. The nuclear vs. cytoplasmic ratio of the HSF2BP proteins in HeLa cells and KGN cells was quantified by ImageJ software.

HeLa cells were cultured and transfected with wild-type or mutant HSF2BP-FLAG plasmids as mentioned above. At 48 h after transfection, the cells were incubated with ETO (5 μg/ml) for 1 h or MMC (500 ng/ml) for 8 h at 37°C to induce DNA damage. Then the culture medium with ETO or MMC was removed, and the cells were cultured with normal medium for a specific time at 37°C. The phosphorylation of the Ser-139 residue of histone variant H2AX (γH2AX; 1:1,000 dilution, Sigma-Aldrich, 05-636), which was a sensitive marker for DNA damage, was analyzed by Western blotting. Three independent experiments were conducted. Moreover, to compare the change of γH2AX level, we quantified the grayscale scores of Western blotting bands using ImageJ software and normalized the levels of γH2AX to the β-actin loading control in each group. The grayscale scores of γH2AX in the cells treated with ETO or MMC and recovered at specific time points were divided by those of untreated cells in each group. We compared the relative grayscale score between the wild-type and mutant groups at specific time points.

Through variation screening in the in-house WES database of sporadic POI, eight variations in HSF2BP (GenBank: NM_007031.2) have been identified. However, only two homozygous missense variations (following recessive inheritance pattern) were found, c.382T > C (p.C128R) and c.557T > C (p.L186P), which were with frequencies below 0.001 in the public databases 1000 Genomes, ExAC, and gnomAD, and were absent in the genomics database of 114,783 normal Han Chinese individuals (PGG.Han, http://www.pgghan.org). Both variants were confirmed by Sanger sequencing (Figure 1B), and the two amino acid sites were highly conserved among species (Figure 1C). According to the ACMG guideline, the two variants were classified as VUS.

Patient POI-1, who carried HSF2BP p.C128R, experienced menarche at 11 years old, underwent irregular menstrual cycle (2–12 months per menstrual cycle) since 13 years old, and suffered amenorrhea at 24 years of age. Patient POI-2, who carried HSF2BP p.L186P, experienced menarche at 16 years old, followed by an irregular menstrual cycle that ceased at the age of 18. Ultrasound examination showed the two patients had atrophic ovaries without follicles (Table 1).

The wild-type or mutant HSF2BP-FLAG plasmids were transfected into HeLa cells and KGN cells, and the immunofluorescence against FLAG was performed to test the cellular localization of protein HSF2BP. The polymorphic variant I93M (rs200533753) was used as the negative control. The wild-type HSF2BP and I93M were expressed in both the cytoplasm and nucleus of HeLa and KGN cells. However, the signals of two mutant proteins (C128R and L186P) localizing in the nucleus were significantly lower than those of the wild-type protein (Figure 2). Moreover, after treatment with ETO or MMC, the localization of mutant proteins in cells was not changed (Figure 3). These results suggested that the variations C128R and L186P impaired the nuclear location of HSF2BP, where its DNA repair function was proposed to be performed, and the disturbed nuclear localization of mutant HSF2BP was not affected by DNA damaging agents.

To elucidate whether the two variations impaired the DNA repair capacity of HSF2BP, DNA damage assays were performed. The result showed that the level of γH2AX in HeLa cells overexpressing wild-type HSF2BP increased immediately after ETO treatment and decreased to the untreated level after recovery for 1 h. Whereas, the γH2AX level in cells overexpressing mutant HSF2BP was significantly higher than that in the wild-type group after recovery for 1 h (Figure 4B). Similar results were observed in the MMC treatment groups, in which the γH2AX level in cells overexpressing mutant HSF2BP was significantly higher than that in wild-type cells after recovery for 6 h (Figure 4C), indicating that the mutant HSF2BP had impaired DNA repair capacity. Considering that HSF2BP was especially expressed in germ cells, the ultimately repaired DNA damage, demonstrated by decreased γH2AX level after 2 h of recovery in ETO-treated cells, might be caused by the endogenous DNA repair mechanism of HeLa cells. Then, with those functional studies, both variations were classified as LP according to the ACMG guideline (Table 1).